Original Article

Proc IMechE Part K:

J Multi-body Dynamics
Lateral control of an autonomous 2019, Vol. 233(1) 141–151
! IMechE 2018

vehicle using integrated backstepping Article reuse guidelines:

sagepub.com/journals-permissions

and sliding mode controller DOI: 10.1177/1464419318797051

journals.sagepub.com/home/pik

Armin Norouzi1 , Milad Masoumi2, Ali Barari3

and Saina Farrokhpour Sani4

Abstract
In this paper, a novel Lyapunov-based robust controller by using meta-heuristic optimization algorithm has been pro-
posed for lateral control of an autonomous vehicle. In the first step, double lane change path has been designed using a
fifth-degree polynomial (quantic) function and dynamic constraints. A lane changing path planning method has been used
to design the double lane change manoeuvre. In the next step, position and orientation errors have been extracted based
on the two-degree-of-freedom vehicle bicycle model. A combination of sliding mode and backstepping controllers has
been used to control the steering in this paper. Overall stability of the combined controller has been analytically proved
by defining a Lyapunov function and based on Lyapunov stability theorem. The proposed controller includes some
constant parameters which have effects on controller performance; therefore, particle swarm optimization algorithm
has been used for finding optimum values of these parameters. The comparing result of the proposed controller with
backstepping controller illustrated the better performance of the proposed controller, especially in the low road fric-
tions. Simulation of designed controllers has been conducted by linking CarSim software with Matlab/Simulink which
provides a nonlinear full vehicle model. The simulation was performed for manoeuvres with different durations and road
frictions. The proposed controller has outperformed the backstepping controller, especially in low frictions.

Keywords
Vehicle lateral control, autonomous vehicle, backstepping controller, sliding mode controller, Lyapunov-based controllers

Date received: 4 April 2018; accepted: 30 July 2018

significant impact on the input and output of the con-

Introduction
trol systems. In vehicle dynamics control, lateral con-
There are many contributing factors to grounded trol of the vehicle is an essential part, because of its
vehicle accidents including poor road condition and role in the stability of the vehicle. In vehicle dynamics
violation of necessary rules such as speeding and espe- control, the yaw stability control system is an essential
cially unsafe lane change. Driver error was cited as part of vehicle stability. Lateral dynamic control of
the underlying cause of from 45% to 75% of fatal vehicles is generally done through braking and steer-
crashes.1 To reduce driver-related accidents, numerous ing subsystems. In yaw control system strategies, the
researches have been conducted to deal with this prob- controller needs to control yaw rate and sideslip angle
lem. Advanced driver assistance systems (ADASs)
have been proposed by transportation experts to
1
address this issue. ADASs can provide autonomous Department of Mechanical Engineering, Faculty of Engineering,
control due to three methods.2 Firstly, the autonomous University of Alberta, Edmonton, Canada
2
Department of Electrical Engineering, Azad University of Garmsar,
vehicle controller obtains surrounding information.
Garmsar, Iran
Next, the controller design desired path based on envir- 3
Department of Mechanical Engineering, K. N. Toosi University of
onmental information. Finally, the controller needs to Technology, Tehran, Iran
4
track the desired path. In this paper, just the control- School of Railway Engineering (SRE), Iran University of Science &
ler’s ability to track the desired path has been studied. Technology, Tehran, Iran
Vehicle dynamics is one of the primary and funda-
Corresponding author:
mental parts of classical mechanics which argues Armin Norouzi, Mechanical Engineering Building, North
about the dynamics of vehicles. Controllers play a Campus,University of Alberta, Edmonton, AB T6G 2G8, Canada.
crucial role in the responding system and have a Email: [email protected]
142 Proc IMechE Part K: J Multi-body Dynamics 233(1)

quickly and responsively to follow the designed path improved the results compared to the classical sliding
by using steering and braking systems.3 Controlling mode controller. In Janbakhsh et al.,15 switching gain
steering has significant impacts on control of the vehi- is updated based on the sliding surface. This control-
cle on the designed path direction, and it is the main ler does not need an upper limit of uncertainty in
discussion of the current paper. determining switching gain as well as the adaptive
Advanced active chassis control systems include sliding surface has a good performance. Based on
four-wheel steering (4WS), active front steering (AFS), the designed controller by Li et al.,16 sliding surface
steer-by-wire (SBW) and direct yaw moment control and its derivations enter a fuzzy logic system and its
(DYC) in which steering system plays the central role. output execute the command to do the manoeuvre. In
Generally, there are three structures in the active steer- FNNs sliding mode controller, sliding surface enters
ing systems that include 4WS, active rear steering (ARS) an FNN controller, and its output generates the vehi-
and active front steering and four-wheel active steering cle command. Results indicated better performance
(4WAS).4 Vehicle active steering plays a crucial role in compare to the sliding mode controller.19 In the adap-
accident avoidance, vehicle handling and vehicle stabil- tive fuzzy sliding mode controller,17 system error
ity when facing changes in road conditions or the pres- enters the fuzzy controller after creating sliding sur-
ence of an obstacle.5 Vehicle active steering has face, and fuzzy controller factors are updated based
applications in tracking the manoeuvres, especially in on the system error. Finally, adding this controller to
autonomous vehicles, and is very useful in tracking the equivalent sliding mode controller provides vehi-
the optimum path. The most common chassis control cle commands. Alipour et al.18 used a combination of
is the 4WS that depends on tire lateral force which is PI controller and sliding mode controller to create an
proportional to the steer angle in a range where the lat- improved sliding mode controller which showed more
eral acceleration is small. Under this circumstance, the precise results compared to the sliding mode controller.
control law can be introduced rather easily by adopting Meta-heuristic optimization algorithms are used enor-
a two-degree-of-freedom (2-DOF) linear vehicle model6 mously to design controllers.20,21 Genetic algorithm
Lyapunov method is a very useful tool for feedback has been used to optimize the FNN controller for
controller design. Many of the feedback control tech- tracking the vehicle’s manoeuvre in lane changing.20
niques are based on the idea of a Lyapunov function Proportional integral derivative (PID) and linear quad-
definition or, more specifically, are based on the deriva- ratic regulator (LQR) controllers are optimized using
tive of the Lyapunov function which guarantees the particle swarm optimization (PSO) algorithm.21 Of
convergence to an equilibrium point or a point of sta- course, one of the fundamental problems in the use
bility. Backstepping and sliding mode controllers are of optimization algorithms is that they take much
feedback controllers based on the Lyapunov. The back- time to respond and convergence which makes them
stepping controller is designed based on the Lyapunov unusable in real-time systems.
function definition. The backstepping controller can use In this study, the backstepping controller is com-
high flexibility to solve stability, tracking and robust bined with a sliding mode controller to design a lateral
control with lower limitations than other methods. controller for tracking the reference path. The PSO
Sliding mode controller is a prevalent method in algorithm is used to optimize control system param-
robust control design. Sliding mode control, the sliding eters. The cost function of this optimization is the
surface gets to an equilibrium point in a limited time and sliding surfaces of the sliding mode controller and
keeps the system in future time on that point.7 the controller output. Based on the references, both
Robust control methods in the design of the lateral backstepping and sliding model controller has good
vehicle controller, particularly control methods based control performance in controlling nonlinear systems.
on Lyapunov stability, are beneficial due to the pres- Therefore, combining these two controllers and using
ence of uncertainties in vehicle systems. Guo et al.8 a meta-heuristic optimization algorithm for finding
used an integrated longitudinal and lateral move- the constant coefficient have achieved the perfect
ments that designed a path-tracking system based on tracking of desired inputs. On the other hand, analyt-
the integral backstepping controller. Another study ical proof of Lyapunov stability for integration con-
designed and tested a backstepping controller for troller guarantees the stability of the proposed
lane change manoeuvre and their results of tests pre- controller. In this study, the two-degree vehicle
sented the proper convergence of errors.9 Unlike model and lateral position and orientation error
backstepping controllers, sliding mode controllers model with constant longitudinal velocity have been
and their integration with other control laws have used. Controller simulation has been conducted using
been extensively used by researchers. Using sliding a full vehicle model by linkage of Matlab/Simulink
mode control based on Lyapunov stability, proper with CarSim. The desired path is a double lane
results of stability and path tracking are presented change manoeuvre which is used for overtaking man-
in literature.10–14 However, its integration with the oeuvres. Simulation results indicated accurate track-
adaptive control law,15 fuzzy systems,16 adaptive ing compare to the backstepping controller in low
fuzzy systems,17 the proportional integral (PI) con- (0.3) and high (0.9) frictions and also in different man-
troller18 and fuzzy neural networks (FNNs)19 oeuvre durations.
Norouzi et al. 143

Path planning in real-time, a bicycle model is used where lateral

There are specified standard lane change manoeuvres accelerations are relatively low.25 Schematic of this
regarding geometry and kinematics as stated in ISO model is shown in Figure 2.
and BS standards, used with full non-linear multi- 8 9 8 9
body models which are more representative of real >
> y>
> >
> 0 > >
>
> >
> >
> >
vehicle dynamics22 but in this study, a five-degree > > > 2Cf > >
<
d y_ = < m
=
polynomial function (Quintic) is used for lane ¼ 
change maneuvers.23 Equation (1) is considered for dt >
> >
> >
> 0 > >
>
> >
> > > >
>
: >
> ; > : 2lf Cf >
;
lane changing path planning. At the start and finish
Iz
time, the vehicle moves in a straight line so the lateral 2 3
acceleration and speed will be equal to zero. Lateral 0 1 0 0
position is equal to zero at the beginning (equation 6 2Cf þ2Cr 2lf Cf 2Cr lr 7
60  0 Vx  7
6 mVx mVx 7
(2)), and the spacing between two lanes is considered þ6 7
60 0 0 1 7
for at the end for it (equation (3)). ti is the start time, 4 5
2lf Cf 2lr Cr 2l2 Cf þ2l2 Cr
and tf is the end time of the manoeuvre. 0  Iz V x 0  f Iz V x r
ð4Þ
yðtÞ ¼ at5 þ bt4 þ ct3 þ dt2 þ et þ f ð1Þ

y_ t¼ti ¼0
 For controlling the vehicle to follow the path,
y€ t¼ti ¼0
 ð2Þ orientation and positioning errors are defined based
y_ t¼tf ¼0 on the distance of the mass centre from the desired
 path and the angle of the desired path from the vehicle
y€ t¼t ¼0
f angle in equation (5).24

yt¼ti ¼ w1 e€1 ¼ y€ þ Vx ð _  Þ, e2 ¼ _  ð5Þ
 d d
ð3Þ
y1 t¼tf ¼ w2
Vehicle error model is obtained by writing vehicle
models based on equation (5)
w1 and w2 are equal to zero and 3.75 for the man- 2 3
8 9 8 9 0
oeuvre, and for the second lane change manoeuvre, > e1 > > 0 >
>
> > > >
> >
they are equal to 3.75 and zero. By applying the d < e_1 = < 2Cf > = 6  2Cf lf 2Cr lr  V 7
6 x7
m mVx
boundary conditions to the candidate equation (1), ¼þ þ6 6
7 _d
7
dt > e
> 2> >
> >
> 0 >
> 4 0 5
the final manoeuvre path is obtained. In this study, >
: ; > >
: 2lf Cf ; 2 2
2Cf lf þ2Cr lr
e_2 Iz  V x Iz
manoeuvres of 3 s duration (tfti ¼ 3) and 5 s dur- 2 3
ation (tfti ¼ 5) are considered for both direct and 0 1 0 0
60  f r 2C þ2C 2C f þ2Cr 2C l
f f þ2C r r 7
l
reverse path. Manoeuvres are shown in Figure 1. 6 mVx m mVx 7
660
7
7
4 0 0 1 5
2Cf lf 2Cr lr 2Cf lf 2Cr lr 2Cf l2f þ2Cr l2r
0  Iz V x Iz  Iz
Modelling 2 3
e1
Vehicle’s model for designing controller 6 e_ 7
6 17
6 7
4 e2 5
The state-space of vehicle 2-DOF bicycle model has
been adopted for designing controller.24 Nevertheless, e_2
ð6Þ

Tyre model
Two-DOF bicycle model has been used for designing
controller. This model has proposed based on linear
tyre model.24 The controller has been applied to
CarSim by linking CarSim software with Matlab/
Simulink. Therefore, the controller has been tested
based on nonlinear tyre mode. The CarSim has been
developed based on several tyre models. The extended
model (more tables for camber effects), the Pacejka
5.2 version of the magic formula, and MF-Tyre
Figure 1. Double lane change designed reference path. from TASS/TNO has been used in running.
144 Proc IMechE Part K: J Multi-body Dynamics 233(1)

Figure 2. A schematic view of a 2-DOF vehicle bicycle model and lateral position and orientation errors.

Controller design Table 1. Nomenclature of the vehicle’s dynamic model.

Preparing the model for controller design Equation Equation
C þC lr Cr lf Cf
For backstepping controller design, the control a11 2 fmV x r a21 2 Iz V x
system should be considered as follows7 C þC lf C f lr Cr
a12 2 f m r a22 2 Iz
l C l C Cf l2f þCr l2r
_ ¼ f ðÞ þ GðÞ ð7Þ a13 2 r rmV xf f a23 2 Iz V x
c l c
b1 2 mf b2 2 fIz f
_ ¼ fa ð, Þ þ Ga ð, Þ u ð8Þ l C l C Cf l2 þCr l2
d1 2 r rmV xf f  Vx d2 2 Ifz V x r
For designing the backstepping controller, equa-
tion (6) should be written as equations (7) and (8); By comparing equations (9) and (10) with equa-
therefore tions (7) and (8), the following equations are obtained
2 3  
x2  ¼ x1 , f ðÞ ¼ 0, GðÞ ¼ 0 1 0
 6 7 2 32 3
x_ 1 ¼ 1 0 0 4 x3 5 ð9Þ a11 a12 a13 x2
x4 6 76 7
fa ¼ 4 0 0 1 5 4 x3 5
2 3 2 32 3 a21 a22 a23 x4
x_ 2 a11 a12
a13 x2 2 3 2 3
6_ 7 6 76 7 b1 x2
4 x3 5 ¼ 4 0 0
1 5 4 x3 5 6 7 6 7
Ga ð, Þ ¼ 4 0 5,  ¼ 4 x3 5
x_ 4 a21 a22 a23 x4
2 3 2 3 ð10Þ b2 x4
b1 d1
6 7 6 7 _ ð12Þ
þ4 0 5 uþ4 0 5 d
b2 d2 Two systems of 2-DOF bicycle model equations
are converted to backstepping controller standardly
X matrix is defined as follows based on Khalil.7

2 3 2 3
x1 e1 Design of backstepping controller
6x 7 6 e_ 7
~ ¼6
X 6 7
27 6 17
¼6 7 ð11Þ The controllers that resist against disturbances
4 x3 5 4 e2 5 and uncertainty of parameters are called robust con-
x4 e_2 trollers. The backstepping controller is a robust con-
troller. Equations (7) and (8) are considered as the
standard equations of the system in which  2 Rn ,
The correlation coefficients of equation (10) are  2 Rm and u 2 Rm (m can be greater than 1).
presented in Table 1. Assume that f, fa , G and Ga are smooth functions
Norouzi et al. 145

(and defined) on the domain of the question, and also Applying equation (18) to equation (9) yields
f and fa are zero at the origin and matrix Ga ðm  mÞ is
non-singular. Also, assume that the system equation x_ 1 ¼ GðÞ ¼ x1 ð20Þ
(equation (8)) can stabilize using a feedback controller
 ¼ ðÞ by ð0Þ ¼ 0 while Lyapunov function To prove the Lyapunov stability, equation (19) is
(smooth and positive defined) VðÞ is applied in the differentiated
following inequality
_
VðÞ ¼ 3x1 x_ 1 ð21Þ
@V
¼ ½ f ðÞ þ GðÞðÞ4  W ðÞ ð13Þ
@ Using equations (20) and (21), the Lyapunov sta-
bility is proved
The Lyapunov candidate function is defined based
on systems equation (equations (7) and (8)) as follow7 _
VðÞ ¼ 3x21 40 ð22Þ

1 Now, considering the system of equations (9) and

Va ¼ VðÞ þ ½  ’ðÞT ½  ’ðÞ ð14Þ
2 (10), and equations (16) and (18), the control input is
obtained as
By differentiating the Lyapunov candidate func- 80 2 31 2 3 0 2 31 9
tion, the desired input is obtained which satisfy the >
> x2 1 0 >
>
>
> B  6 7C 6 7 B 6 7C > >
stability theorem of Lyapunov condition >
> B 0 1 0 6 7 C 6 7 B 6 7 C >
>
> @ x
4 3 5A 4 1 
5 @ 3x 1 4 5A >
1 >
>
>
> >
>
>
< >
=
@V @V x4 1 0
V_ a ¼ ¼ ð f þ GÞ þ Cð  Þ uB ¼ GTa
>
2 32 3 2 3
>
@ @ >
> a a12 a13 x2 x2 þ x1 >
>
  ð15Þ > 6 11
> 76 7 6 7 >
>
@ >
> >
>
þ ½  T fa þ Ga u  ð f þ GÞ >
> 6
4 0 0 7 6 7 6
1 5 4 x3 5  k4 x3 þ x1 57 >
>
>
> >
>
@ : ;
a21 a22 a23 x4 x4 þ x1
Considering equation (15), the control law is ð23Þ
obtained as
" # Simplifying equation (23), the control law of back-
 T stepping controller is obtained
@ @V
u ¼ G1
a ð f þ GÞ  G  f  kð  Þ
@ @
u^ B ¼ðkðb1 þ b2 ÞÞ e1  ðb1 þ b2 þ b1 a12 þ b2 a22 Þ e2
 ðb1 a11 þ b2 a21 þ b1 kÞ e_1  ðb1 a13 þ b2 a23 þ b2 kÞ e_2
k40 ð24Þ
ð16Þ
u^ B
Applying the above control law, the stability of the uB ¼ ð25Þ
ðb21 þ b22 Þ
system is proved using the Lyapunov stability
theorem. where uB is the control input yielded from backstep-
ping controller. The value of k is considered so large
@V that can be robust against uncertainties.
V_ a ¼ ð f þ GÞ  kð  ÞT ð  ’Þ
@ ð17Þ
4WðÞ  kð  ÞT ð  Þ Designing the backstepping controller integrated
 ¼ 0 with sliding mode controller
Equation (17) indicates that the origin
( ¼ 0,  ¼ 0) has asymptotic stability. Similar to the backstepping controller, based on sta-
Using the system obtained in equations (9) and bilizing of  ¼ ðÞ provided that ð0Þ ¼ 0. We are
(10), and employing equation (11), the backstepping also pursuing the function VðÞ, such that for any
controller could be applied to 2-DOF model of vehi- vð, Þ 2 D and constant c thus
cle. The feedback control law stabilizer and the
Lyapunov function VðÞ is defined as _
V4c kk22 ð26Þ
2
3
x1 The above equation indicates that =0 is the
6 7 asymptotic stable equilibrium point of the following
 ¼  ¼ 4 x1 5 ð18Þ
system
x1
_ ¼ f ðÞ þ GðÞ ðÞ ð27Þ
1 3
VðÞ ¼ T  ¼ x21 ð19Þ
2 2
146 Proc IMechE Part K: J Multi-body Dynamics 233(1)

The sliding surface is defined as equation (28), and global best position known so far. These pos-
which is, in fact, the difference between  and ’ðÞ. itions are updated each time a new position is
found. Therefore, one expects for the particles to
S ¼   ’ðÞ ð28Þ move towards the best possible solution.
In order to prove that the slip surface is converged
In addition, the Lyapunov candidate function is towards zero, the Lyapunov function is considered as
considered as equation (29) follows. Through differentiating the equations, the
stability is proved analytically.
1
Va ð, Þ ¼ VðÞ þ ðSÞ2 ð29Þ
2 1
Vs ¼ S2
2  
So, the derivatives of Va in direction of equations @’
¼ SS_ ¼ S _  _
(7) and (8) is obtained as @
  ð34Þ
  @’ðÞ
@’ ¼ S fa þ Ga u  ð f þ GÞ
V_ a ¼ V_ þ ð  ’Þ fa þ Ga u  ð f þ GÞ ð30Þ @
@
4Sðk sgnðSÞÞ
and the control input is as 4  kjSj
 
@ @V By choosing a k greater than zero, the sliding sur-
u ¼ G1
a ð f þ GÞ  G  fa  k sgnð  ’Þ
@ @ face is converged towards zero.
k50 ð31Þ In order to design the controller for the bicycle
model of vehicle, the  ¼ ðÞ applied on the back-
stepping control (equation (18)) is employed. Using
Using equations (26) and (31), the designed con- equation (28), the sliding surface is obtained.
troller’s stability (equation (30)) can be proved using 2 3 2 3 2 3 2 3
the Lyapunov stability theorem as s1 1 x2 x2 þ x1
  6 7 6 7 6 7 6 7
@’ S¼6 7 6 7 6 7 6 7
4 s2 5 ¼   ’ðÞ ¼ x1 4 1 5 þ 4 x3 5 ¼ 4 x3 þ x1 5
_ _ 2
Va ¼ V þ SS4  ckk2 þS   _ _
@ s3 1 x4 x4 þ x1
 
2 @’ ð35Þ
4  ckk2 þS fa þ Ga u  ð f  GÞ 
@
4  ckk22 þsðk sgnðSÞÞ Considering the matrix Ga, the sliding surface s2 is
zero, because the second row of Ga matrix is zero.
4  ckk22 þðkÞ jSj Thus, it has no effect on the control input. The
ð32Þ input control of the integrated controller is obtained
using equation (31).
Choosing k50 thus 22 3 0 2 31 02 313
  1 x2 0
V_ a 4  kk22 þjSj ð33Þ 66 B  6 7C B6 7C7
6 4 1 7
5 @ 0 1 0 4 x3 5A  @4 3x1 5 A 7
6 7
where is positive. So, according to Lyapunov stabil- 6 7
T6
6 1 x4 0 7
ity theorem, this system is always stable. In the pres- uBS ¼ Ga 6 2 32 3 2 3 7
a a a x sgnðs Þ 7
ence of disturbances, the condition k for the stability 6 11 12 13 2 1 7
6 6 76 7 6 7 7
of both controllers (i.e. backstepping controller and 6 4 0 0 1 5 4 x 5  k 4 sgnðs Þ 5 7
4 3 2 5
backstepping controller integrated with sliding mode a21 a22 a23 x4 sgnðs3 Þ
controller) must be greater than upper bounds of the
ð36Þ
disturbance. Choosing a greater k, one would resist
against disturbances, but the control input would where uBS is the control input of the proposed con-
have also high value. In this paper, advantages of troller. Simplifying equation (37), the control input is
defining an appropriate cost function and applying obtained as
the PSO meta-heuristic method were taken. PSO is
an iterative computing method which optimizes a u^ B,S ¼ ðb1 a12 þ b2 a22 þ b1 þ b2 Þ x3
problem through improving possible solutions in
terms of certain quality criteria. Considering an initial  ðb1 a13 þ b2 a23 Þ x4  ðb1 a11 þ b2 a21 Þ x2
population of possible solutions as particles and  k ðb1 sgn ðx2 þ x1 Þ þ b2 sgn ðx4 þ x1 ÞÞ
moving them in search space, the PSO method opti- u^ B:S
mizes and solves the problem using a simple formula uB:S ¼ 2
ðb1 þ b22 Þ
in terms of the position and velocity of each particle.
ð37Þ
The movements of particles are influenced by the local
Norouzi et al. 147

Taking into account the chattering phenomenon in is the F-class model with specifications summarized in
the sliding mode controller, the Sat function has been Table 2. Simulations were conducted for two double
used rather than the Sign function. Noting that the lane change manoeuvres of 3 and 5 s. To demonstrate
value of s2 is equal to zero, the sliding surfaces are the robustness of the designed controllers, the simu-
redefined as lation was done for low (0.3) and high (0.9) frictions
and a longitudinal velocity of 30 m/s. Figures 4 and 5
s1 ¼ x2 þ x1
ð38Þ
s2 ¼ x4 þ x1
Table 2. The vehicle parameters in this study.
ðb1 a12 þ b2 a22 Þ ðb1 a13 þ b2 a23 Þ
uB:S ¼ 2 2
e_1  e_2 Symbol Description Quantity
b1 þ b2 b21 þ b22
ðb1 a12 þ b2 a22 þ b1 þ b2 Þ m Mass 1704.7 kg
 e2 Iz Yaw moment of inertia 3048.1 kg.m2
b21 þ b22

lf Front axle-COG distance 1.035 m
k b1 satðsl1 Þ þ b2 satðsl2 Þ
 lr Rear axle-COG distance 1.655 m
b21 þ b22
Cf Cornering stiffness of front tire 105850 N/rad
ð39Þ Cr Cornering stiffness of the rear tire 79030 N/rad
 Road friction coefficient [0.3–0.9]
where l indicates the thickness of the boundary layer. Vx Longitude velocity 30 m/s
The k value in equation (39) is obtained using the PSO ti lane change manoeuvres start time
method and the cost function shown in equation (40). tf lane change manoeuvres end time
The cost function is defined in terms of minimizing the  Front wheel steering angle
sliding surface and control input. d Desired yaw rate from road
Z t
e1 Lateral position error concern-
Costfunc: ¼ ðjs1 j þ js2 j þ jujÞ dt ð40Þ ing the road
0 e2 Yaw angle error concerning
the road
The controlling diagram of the proposed control
Thickness of the boundary layer
law is shown in Figure 3.
K Switching gain of proposed
controller
Simulation and results  Constant Value in Lyapunov
stability proof
In this paper, the Simulink and CarSim software have
S Sliding surface matrix
been linked together to simulate the system using the
V Lyapunov function
full vehicle model. The equations of motion in the
uB Controller output of backstepping
CarSim math models are valid for full nonlinear 3D
controller
motions of rigid bodies. The components that have
uB:S Controller output of backstep-
significant effect on handling, braking, and acceler-
ping-sliding surface based on
ation are represented with nonlinear tables of meas- PSO optimization controller
urable data. The used vehicle model for the simulation

Figure 3. The controlling diagram of the proposed integrated control.

148 Proc IMechE Part K: J Multi-body Dynamics 233(1)

Figure 4. The simulation of backstepping controller and the proposed controller for manoeuvre of 5 s and friction of 0.9.

Figure 5. The simulation of backstepping controller and the proposed controller for manoeuvre of 5 s and friction of 0.3.

illustrate the simulation results for the manoeuvre of (control output), the yaw angle, slip angle and the roll
5 s. Figures 6 and 7 illustrate the simulation results for angle of the vehicle, respectively.
the manoeuvre of 3 s. Sub-figures from left to right For higher frictions and different manoeuvres,
show simulated and desired vehicle paths, sliding sur- the performance of both controller are almost the
face of the proposed controller, position and orienta- same. However, the performance of the proposed con-
tion errors with respect to time, steering of the vehicle troller in eliminating the position error is remarkable.
Norouzi et al. 149

Figure 6. The simulation of backstepping controller and the proposed controller for manoeuvre of 3 s and friction of 0.9.

Figure 7. The simulation of backstepping controller and the proposed controller for manoeuvre of 3 s and friction of 0.3.

In terms of the steering angle, the proposed controller achieving the full tracking. In terms of the roll
has lower effort, but associated with it some variations angle, the proposed controller shows better results
which should be addressed in the future works. For again. For manoeuvre of 3 s, the controller has pre-
lower frictions, the performance of the proposed con- served its good performance which results in low roll
troller is evidently significant. For manoeuvre of 5 s, slip angles. The variations of the steering in this man-
the controller shows better results, particularly in oeuvre are significant, and their effect on the path is
minimizing the position and orientation errors and evident. As it is shown in Figure 7(a), the vehicle has
150 Proc IMechE Part K: J Multi-body Dynamics 233(1)

some low-amplitude variations about its desired path, Funding

which are results of the variations in the performance The author(s) received no financial support for the research,
of the controller. In all manoeuvres and under any authorship, and/or publication of this article.
condition, the sliding surface has approached towards
zero and the vehicle has shown an acceptable stability. ORCID iD
This indicates that our novel integrated controller has Armin Norouzi http://orcid.org/0000-0003-2690-0739
good performance.
References
Conclusion 1. Wierwille WW, Hanowski RJ, Hankey JM, et al.
Identification and evaluation of driver errors: overview
In this paper, using a fifth degree polynomial and recommendations. Report No. FHWA-RD-02-003,
function and dynamical constraints of vehicle, the 2002.
double lane change path for two manoeuvre length 2. Urmson C, Anhalt J, Bagnell D, et al. Autonomous
of 3 and 5 s has been designed. A novel controller driving in urban environments: boss and the urban chal-
which utilized combination/integration of a backstep- lenge. J Field Robot 2008; 25: 425–466.
ping and sliding mode controllers have been pro- 3. Kemao P, Sam Y, Ismail M, et al. A review on inte-
posed. The stability of the proposed controller has grated active steering and braking control for vehicle
been proved by defining a suitable Lyapunov function yaw stability system. Jurnal Teknologi (Science &
Engineering) 2014. eISSN: 2180–3722.
and proving its stability analysis. There are some con-
4. Li B and Fu Z-J. Coordinated control of active steering
stant parameters in the design of every controller and active roll control systems for enhanced vehicle lat-
which finding the optimum value of these parameters eral dynamics. Int J Veh Perform 2016; 2: 418–434.
is a time demanding process. Therefore, the proposed 5. Park J, Kim D, Yoon Y, et al. Obstacle avoidance of
controller’s parameter has been optimized by using autonomous vehicles based on model predictive control.
the PSO method. The defined cost function is a sum Proc IMechE, Part D: J Automobile Engineering 2009;
of absolute values of sliding surfaces and control 223: 1499–1516.
effort. The simulation of this study has been con- 6. Abe M. Vehicle dynamics and control for improving
ducted by linking the Simulink to CarSim software handling and active safety: From four-wheel steering
which results in a non-linear full vehicle model with to direct yaw moment control. Proc IMechE, Part K:
steering and suspension subsystems. This controller J Multi-body Dynamics 1999; 213: 87–101.
7. Khalil HK. Nonlinear systems. Upper Saddle River:
has been simulated for different frictions. The results
Prentice-Hall, 1996.
of the simulation indicated proper tracking of the 8. Guo L, Ge P-S, Yue M, et al. Lane changing trajectory
proposed controller compared to the backstepping planning and tracking controller design for intelligent
controller. vehicle running on curved road. Math Probl Eng 2014;
The summary of this study achievement is as 2014: 478573.
follows: 9. Feng Y, Rongben W and Ronghui Z. Algorithm on
lane changing and tracking control technology for intel-
1. Stability analysis of an integrated sliding mode ligent vehicle. In: 2007 IEEE international conference on
controller and backstepping controller has been robotics and biomimetics (ROBIO), 2007, pp.1888–
proved analytically. 1893. Piscataway: IEEE.
2. Integrated sliding mode controller and backstep- 10. Akar M and Kalkkuhl JC. Lateral dynamics emulation
via a four-wheel steering vehicle. Veh Sys Dyn 2008; 46:
ping controller has better performance versus to
803–829.
sliding mode controller especially in the lower 11. Hiraoka T, Nishihara O and Kumamoto H. Automatic
bound of uncertainty. path-tracking controller of a four-wheel steering vehi-
3. Adding a meta-heuristic optimization algorithm cle. Veh Sys Dyn 2009; 47: 1205–1227.
for finding the constant coefficients of the pro- 12. He J, Crolla DA, Levesley M, et al. Coordination of
posed controller is helpful to reduce tracking active steering, driveline, and braking for integrated
error. vehicle dynamics control. Proc IMechE, Part D: J
4. Using meta-heuristic optimization increase the Automobile Engineering 2006; 220: 1401–1420.
runtime of the system. 13. Kunnappillil Madhusudhanan A, Corno M and
Holweg E. Sliding mode-based lateral vehicle dynamics
control using tyre force measurements. Veh Sys Dyn
Reducing the running time of the proposed control
2015; 53: 1599–1619.
law, and real-time tests for proposed controller will be
14. Lei J, Wu H, Yang J, et al. Sliding mode lane keeping
introduced as future works. control based on separation of translation and rotation
movement. Optik Int J Light Elec Opt 2016; 127:
Declaration of Conflicting Interests 4369–4374.
The author(s) declared no potential conflicts of interest with 15. Janbakhsh AA, Bayani Khaknejad M and Kazemi R.
respect to the research, authorship, and/or publication of Simultaneous vehicle-handling and path-tracking
this article. improvement using adaptive dynamic surface control
Norouzi et al. 151

via a steer-by-wire system. Proc IMechE, Part D: J 21. Salehpour S, Pourasad Y and Taheri SH. Vehicle path
Automobile Engineering 2013; 227: 345–360. tracking by integrated chassis control. J Central South
16. Li L, Lian J, Wang M, et al. Fuzzy sliding mode lateral Univ 2015; 22: 1378–1388.
control of intelligent vehicle based on vision. Adv Mech 22. Hegazy S, Rahnejat H and Hussain K. Multi-body
Eng 2013; 5: 216862. dynamics in full-vehicle handling analysis. Proc
17. Guo J, Li L, Li K, et al. An adaptive fuzzy- IMechE, Part K: J Multi-body Dynamics 1999; 213:
sliding lateral control strategy of automated vehicles 19–31.
based on vision navigation. Veh Sys Dyn 2013; 51: 23. Norouzi A, Kazemi R and Azadi S. Vehicle lateral con-
1502–1517. trol in the presence of uncertainty for lane change man-
18. Alipour H, Bannae Sharifian MB and Sabahi M. A euver using adaptive sliding mode control with fuzzy
modified integral sliding mode control to lateral stabil- boundary layer. Proc IMechE, Part I: J Systems and
isation of 4-wheel independent drive electric vehicles. Control Engineering 2018; 232: 12–28.
Veh Sys Dyn 2014; 52: 1584–1606. 24. Rajamani R. Vehicle dynamics and control. Berlin,
19. Li L, Wang H, Lian J, et al. A lateral control method of Germany: Springer Science & Business Media, 2011.
intelligent vehicle based on fuzzy neural network. Adv 25. Segel L. Theoretical prediction and experimental sub-
Mech Eng 2014. stantiation of the response of the automobile to steering
20. Feng J, Ruan J and Li Y. Study on intelligent vehicle control. Proc Inst Mech Eng Autom Div 1956; 10:
lane change path planning and control simulation. In: 310–330.
2006 IEEE international conference on information
acquisition, 2006, pp.683–688. Piscataway: IEEE.